Hi,
> > For the ipvsadm option, I've used (-b|--sched-flags) 123. I don't
> Not sure if we need "-b".
Well, okay. I just have a personal preference for short options. :-)
It makes ipvsadm -Sn output look nicer, too.
> I guess it is difficult to maintain many options,
> may be one option --sched-flags should be enough, for example:
> --sched-flags sh-fallback,sh-port
Okay, that looks good.
I've done some refactoring to simplify ip_vs_sh_schedule. Let me know
what you think.
diff --git a/include/uapi/linux/ip_vs.h b/include/uapi/linux/ip_vs.h
index a245377..2945822 100644
--- a/include/uapi/linux/ip_vs.h
+++ b/include/uapi/linux/ip_vs.h
@@ -20,6 +20,12 @@
#define IP_VS_SVC_F_PERSISTENT 0x0001 /* persistent port */
#define IP_VS_SVC_F_HASHED 0x0002 /* hashed entry */
#define IP_VS_SVC_F_ONEPACKET 0x0004 /* one-packet scheduling */
+#define IP_VS_SVC_F_SCHED1 0x0008 /* scheduler flag 1 */
+#define IP_VS_SVC_F_SCHED2 0x0010 /* scheduler flag 2 */
+#define IP_VS_SVC_F_SCHED3 0x0020 /* scheduler flag 3 */
+
+#define IP_VS_SVC_F_SCHED_SH_FALLBACK IP_VS_SVC_F_SCHED1 /* SH fallback */
+#define IP_VS_SVC_F_SCHED_SH_PORT IP_VS_SVC_F_SCHED2 /* SH use port */
/*
* Destination Server Flags
diff --git a/net/netfilter/ipvs/ip_vs_sh.c b/net/netfilter/ipvs/ip_vs_sh.c
index 0df269d..4bb9636 100644
--- a/net/netfilter/ipvs/ip_vs_sh.c
+++ b/net/netfilter/ipvs/ip_vs_sh.c
@@ -48,6 +48,10 @@
#include <net/ip_vs.h>
+#include <net/tcp.h>
+#include <linux/udp.h>
+#include <linux/sctp.h>
+
/*
* IPVS SH bucket
@@ -71,10 +75,37 @@ struct ip_vs_sh_state {
struct rcu_head rcu_head;
};
+
+/*
+ * If the dest flags is set with IP_VS_DEST_F_OVERLOAD,
+ * consider that the server is overloaded here.
+ */
+static inline int is_overloaded(struct ip_vs_dest *dest)
+{
+ return dest->flags & IP_VS_DEST_F_OVERLOAD;
+}
+
+
+/*
+ * Helper function to determine if server is unavailable
+ */
+static inline int
+is_unavailable(struct ip_vs_dest *dest)
+{
+ return (!dest ||
+ atomic_read(&dest->weight) <= 0 ||
+ is_overloaded(dest));
+}
+
+
/*
* Returns hash value for IPVS SH entry
*/
-static inline unsigned int ip_vs_sh_hashkey(int af, const union nf_inet_addr *addr)
+static inline unsigned int
+ip_vs_sh_hashkey(int af,
+ const union nf_inet_addr *addr,
+ __be16 port,
+ unsigned int offset)
{
__be32 addr_fold = addr->ip;
@@ -83,7 +114,8 @@ static inline unsigned int ip_vs_sh_hashkey(int af, const union nf_inet_addr *ad
addr_fold = addr->ip6[0]^addr->ip6[1]^
addr->ip6[2]^addr->ip6[3];
#endif
- return (ntohl(addr_fold)*2654435761UL) & IP_VS_SH_TAB_MASK;
+ return (offset + (ntohs(port) + ntohl(addr_fold))*2654435761UL) &
+ IP_VS_SH_TAB_MASK;
}
@@ -91,13 +123,55 @@ static inline unsigned int ip_vs_sh_hashkey(int af, const union nf_inet_addr *ad
* Get ip_vs_dest associated with supplied parameters.
*/
static inline struct ip_vs_dest *
-ip_vs_sh_get(int af, struct ip_vs_sh_state *s, const union nf_inet_addr *addr)
+ip_vs_sh_get(struct ip_vs_service *svc,
+ struct ip_vs_sh_state *s,
+ const union nf_inet_addr *addr,
+ __be16 port)
{
- return rcu_dereference(s->buckets[ip_vs_sh_hashkey(af, addr)].dest);
+ struct ip_vs_dest *dest;
+
+ dest = rcu_dereference(
+ s->buckets[ip_vs_sh_hashkey(svc->af, addr, port, 0)].dest);
+
+ return is_unavailable(dest) ? NULL : dest;
}
/*
+ * As ip_vs_sh_get, but with fallback if selected server is unavailable
+ */
+static inline struct ip_vs_dest *
+ip_vs_sh_get_fallback(struct ip_vs_service *svc,
+ struct ip_vs_sh_state *s,
+ const union nf_inet_addr *addr,
+ __be16 port)
+{
+ unsigned int offset;
+ struct ip_vs_dest *dest;
+
+ for (offset = 0; offset < IP_VS_SH_TAB_SIZE; offset++) {
+ dest = rcu_dereference(s->buckets[
+ ip_vs_sh_hashkey(svc->af, addr, port, offset)].dest);
+ if (!is_unavailable(dest))
+ return dest;
+#ifdef CONFIG_IP_VS_DEBUG
+ else if (dest)
+ IP_VS_DBG_BUF(6, "SH: selected unavailable server "
+ "%s:%d (offset %d)",
+ IP_VS_DBG_ADDR(svc->af, &dest->addr),
+ ntohs(dest->port),
+ offset);
+ else
+ IP_VS_DBG(6, "SH: selected null server "
+ "(offset %d)",
+ offset);
+#endif
+ }
+
+ return NULL;
+}
+
+/*
* Assign all the hash buckets of the specified table with the service.
*/
static int
@@ -214,12 +288,34 @@ static int ip_vs_sh_dest_changed(struct ip_vs_service *svc,
/*
- * If the dest flags is set with IP_VS_DEST_F_OVERLOAD,
- * consider that the server is overloaded here.
+ * Helper function to get port number
*/
-static inline int is_overloaded(struct ip_vs_dest *dest)
+static inline __be16
+ip_vs_sh_get_port(const struct sk_buff *skb, struct ip_vs_iphdr *iph)
{
- return dest->flags & IP_VS_DEST_F_OVERLOAD;
+ __be16 port;
+ struct tcphdr _tcph, *th;
+ struct udphdr _udph, *uh;
+ sctp_sctphdr_t _sctph, *sh;
+
+ switch (iph->protocol) {
+ case IPPROTO_TCP:
+ th = skb_header_pointer(skb, iph->len, sizeof(_tcph), &_tcph);
+ port = th->source;
+ break;
+ case IPPROTO_UDP:
+ uh = skb_header_pointer(skb, iph->len, sizeof(_udph), &_udph);
+ port = uh->source;
+ break;
+ case IPPROTO_SCTP:
+ sh = skb_header_pointer(skb, iph->len, sizeof(_sctph), &_sctph);
+ port = sh->source;
+ break;
+ default:
+ port = 0;
+ }
+
+ return port;
}
@@ -232,17 +328,23 @@ ip_vs_sh_schedule(struct ip_vs_service *svc, const struct sk_buff *skb)
struct ip_vs_dest *dest;
struct ip_vs_sh_state *s;
struct ip_vs_iphdr iph;
-
- ip_vs_fill_iph_addr_only(svc->af, skb, &iph);
+ __be16 port = 0;
IP_VS_DBG(6, "ip_vs_sh_schedule(): Scheduling...\n");
+ ip_vs_fill_iph_skb(svc->af, skb, &iph);
+
+ if (svc->flags & IP_VS_SVC_F_SCHED_SH_PORT)
+ port = ip_vs_sh_get_port(skb, &iph);
+
s = (struct ip_vs_sh_state *) svc->sched_data;
- dest = ip_vs_sh_get(svc->af, s, &iph.saddr);
- if (!dest
- || !(dest->flags & IP_VS_DEST_F_AVAILABLE)
- || atomic_read(&dest->weight) <= 0
- || is_overloaded(dest)) {
+
+ if (svc->flags & IP_VS_SVC_F_SCHED_SH_FALLBACK)
+ dest = ip_vs_sh_get_fallback(svc, s, &iph.saddr, port);
+ else
+ dest = ip_vs_sh_get(svc, s, &iph.saddr, port);
+
+ if (!dest) {
ip_vs_scheduler_err(svc, "no destination available");
return NULL;
}
Alex
--
To unsubscribe from this list: send the line "unsubscribe lvs-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
